yes, it's possible, just look at D3DXMATRIX from DirectX SDK

D3DXMATRIX m;

m(0,0) = 1.0f;

tomek_zielinski:

It's not quite that simple ;). What Tom is talking about is how to register the indexing operator with AngelScript, not just accessing the elements.

zola:

Well, it would certainly be possible to implement an overloaded operator that take two parameters like this. It shouldn't be much more difficult either. I'll have to think about it, because once implemented I can't remove it again since I try to keep backwards compatibility as much as possible.

The way to implement it today is to let the first indexing operator return a pointer to a matrix row (new registered type), which in turn has an indexing operator to access the elements of that row.

If you register the matrix row type with a size of 0 you prevent the scripts from declaring variables of that type. For C++ the pointer to the matrix row is just a pointer to the first element in the matrix.

Access this way wouldn't be very fast though (two system function calls), but at least the syntax would be what you want.

Oh, sorry - my mistake - I didn't see that it was about AngelScript. Sorry once again...

zola:

You could also at least improve your set() and get() pair by a method float &m(int r, int c); which would allow your scripts to write the following:

matrix mtx;

mtx.m(1,1) = 32;
mtx.m(2,1) = 32;

Still not the correct way, but it is closer, don't you think?

I'll think some more about this. After all, C++ has the possibility of overloading the () to make a function call directly. I'll just have to decide if I want AngelScript to have this operator overload or not. The syntax for registering the operator would be:

engine->RegisterTypeBehaviour("matrix", asBEHAVE_INDEX, "float &f(int r, int c)", asMETHOD(matrix, m), asCALL_THISCALL);

Using this from the script would be like

matrix mtx;
mtx[r][c] = 1;

But as I said, I'll decide if this is something AngelScript will have (though it probably will as I don't see any real reason not to :))

Just thought of something that prevents me from implementing multilevel indices like this.

Suppose we register two index operators [](int r), and [](int r, int c), where the first one returns an object that also has an index operator, e.g. matrix row. This would create an ambiguity for the compiler, that could only be solved with parenthesis.

For now, I think I'll keep it the way it is. At least until native support for arrays are available.

The best implementation would then be:

engine->RegisterObjectType("matrix4", sizeof(matrix4), 0);engine->RegisterObjectType("matrix4_row", 0, 0);engine->RegisterTypeBehaviour("matrix4", asBEHAVE_INDEX, "matrix4_row &f(int)", asFUNCTION(matrix4_row), asCALL_CDECL_OBJLAST);engine->RegisterTypeBehaviour("matrix4_row", asBEHAVE_INDEX, "float &f(int)", asFUNCTION(matrix4_col), asCALL_CDECL_OBJLAST);

float &matrix4_row(int r, matrix4 &mtx){  if( r < 0 || r > 3 )     // set exception on the active context  return &mtx[r][0];}float &matrix4_col(int c, float *col){  if( c < 0 || c > 3 )     // set exception on the active context  return &col[c];}

I think the above code should work. But I haven't tried it, nor compiled it so you should only use it as a guide for your own implementation.

Regards,
Andreas